Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?
This branch is 4 commits ahead, 386 commits behind comtihon:master.

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

This is the MongoDB driver for Erlang. MongoDB is a document-oriented database management system. A driver is a client library that provides an API for connecting to MongoDB servers, performing queries and updates on those servers, and performing administrative tasks like creating indexes and viewing statistics.

This version of the driver supports connecting to a single server or replica set, and pooling of both types of connections. Both connection types and pools are thread-safe, i.e. multiple processes can use the same connection/pool simultaneously without interfering with each other.

This driver is implemented as an Erlang application named mongodb. It depends on another Erlang library application named bson, which defines the document type and its standard binary representation. You need both of these. Below we describe the mongodb application; you should also see the bson application to understand the document type.


Download and compile the application

$ git clone git:// mongodb
$ cd mongodb
$ make

Then you have two choices, either install them in your standard Erlang library location. To find the standard location, perform the following in erl:

$ erl
1> code:lib_dir().

To copy the libraries:

$ mkdir <path from above>/mongodb-master
$ cp -R deps ebin <path from above>/mongodb-master/

Or alternatively you can include in your path on startup:

$ erl -pa ebin deps/*/ebin


The mongodb application needs be started before using (to initialize an internal ets table of counters)

> application:start (mongodb).

Although the mongodb application includes several modules, you should only need to use mongo, which is the top-level interface for the driver, and the generic pool if desired. Likewise, you should only need to use the bson module in the bson application.


To connect to a mongodb server listening on localhost:27017 (or any address & port of your choosing)

> Host = {localhost, 27017}.
> {ok, Conn} = mongo:connect (Host).

27017 is the default port so you could elide it in this case and just supply localhost as the argument. mongo:connect returns {error, Reason} if it failed to connect.

Remember to close the connection when finished using it via mongo:disconnect.

To connect to a replica set named "rs1" with seed list of members: localhost:27017 & localhost:27018

> Replset = {<<"rs1">>, [{localhost, 27017}, {localhost, 27018}]}.
> Conn = mongo:rs_connect (Replset).

mongo:do below will connect to the primary or a secondary in the replica set depending on the read-mode supplied.


A database operation happens in the context of a connection (single server or replica set), database, read-mode, and write-mode. These four parameters are supplied once at the beginning of a sequence of read/write operations. Furthermore, if one of the operations fails no further operations in the sequence are executed and an error is returned for the entire sequence.

> mongo:do (safe, master, Conn, test, fun() ->
	mongo:delete (foo, {}),
	mongo:insert (foo, {x,1, y,2}),
	mongo:find (foo, {x,1}) end).

safe, along with {safe, GetLastErrorParams} and unsafe, are write-modes. Safe mode makes a getLastError request after every write in the sequence. If the reply says it failed then the rest of the sequence is aborted and mongo:do returns {failure, {write_failure, Reason}}, or {failure, not_master} when connected to a slave. An example write failure is attempting to insert a duplicate key that is indexed to be unique. Alternatively, unsafe mode issues every write without a confirmation, so if a write fails you won't know about it and remaining operations will be executed. This is unsafe but faster because you there is no round-trip delay.

master, along with slave_ok, are read-modes. master means every query in the sequence must read fresh data (from a master/primary server). If the connected server is not a master then the first read will fail, the remaining operations will be aborted, and mongo:do will return {failure, not_master}. slave_ok means every query is allowed to read stale data from a slave/secondary (fresh data from a master is fine too).

Conn is the connection or rs_connection we send the operations to. If the connection fails during one of the operations then the remaining operations are aborted and {failure, {connection_failure, Reason}} is returned.

test is the name of the database in this example. Collections accessed in the sequence (foo in this example) are taken from the database given in this fourth argument. If a collection is missing from the database it will be automatically created upon first access.

If there are no errors in the sequence of operations then the result of the last operation is returned as the result of the entire mongo:do command. It is wrapped in an ok tuple as in {ok, Result} to distinguish it from an error.

mongo:find returns a cursor holding the pending list of results, which are accessed using mongo:next to get the next result, and mongo:rest to get the remaining results. Either one throws {cursor_expired, Cursor} if the cursor was idle for more than 10 minutes. This exception is caught by mongo:do and returned as {failure, {cursor_expired, Cursor}}. mongo:rest also closes the cursor, otherwise you should close the cursor when finished using mongo:close_cursor.

See the mongo module for a description of all operations. A type specification is provided with each operation so you know the expected arguments and results. The spec line also has a comment if it performs a side-effect such as IO and what exceptions it may throw. No comment means it is a pure function. Also, see the bson module in the bson application for details on the document type and its value types.


This driver does not provide helper functions for commands. Use mongo:command directly and refer to the MongoDB documentation for how to issue raw commands.

There are functions for complex commands like mongo:auth, mongo:add_user, and mongo:create_index.


To authenticate use function mongo:auth.

Plain Erlang string is interpreted as a BSON array of integers, so make sure to always encode your strings, as in <<"hello">> or bson:utf8("hello").


A single (replset-)connection is thread-safe, i.e. multiple mongo:do actions can access it simultaneously. However, if you want to increase concurrency by using multiple connection simultaneously, you can create a pool of connections using the generic resource_pool module with the appropriate factory object supplied by mongo module.

To create a connection pool of max size 10 to a single Host

> Pool = resource_pool:new (mongo:connect_factory (Host), 10).

To create a rs-connection pool of max size 10 to a Replset

> Pool = resource_pool:new (mongo:rs_connect_factory (Replset), 10).

To get a (replset-)connection from the pool

> {ok, Conn} = resource_pool:get (Pool).

Conn can then be supplied to mongo:do. resource_pool:get will return {error, Reason} if can't connect.

Close the pool when done using it and all its connections via resource_pool:close.

More Documentation

API Docs - Documentation generated from source code comments


MongoDB driver for Erlang






No packages published


  • Erlang 97.7%
  • Makefile 2.3%